diff --git a/tdigest/include/tdigest.hpp b/tdigest/include/tdigest.hpp
index 2d3620b1..095752e9 100644
--- a/tdigest/include/tdigest.hpp
+++ b/tdigest/include/tdigest.hpp
@@ -316,7 +316,7 @@ template<typename T, typename A>
 class tdigest<T, A>::const_iterator {
 public:
   using iterator_category = std::input_iterator_tag;
-  using value_type = std::pair<const T&, const W>;
+  using value_type = std::pair<T, W>;
   using difference_type = void;
   using pointer = const return_value_holder<value_type>;
   using reference = const value_type;
diff --git a/tdigest/test/CMakeLists.txt b/tdigest/test/CMakeLists.txt
index 18bf3599..8dcfb4f0 100644
--- a/tdigest/test/CMakeLists.txt
+++ b/tdigest/test/CMakeLists.txt
@@ -39,6 +39,7 @@ target_sources(tdigest_test
   PRIVATE
     tdigest_test.cpp
     tdigest_custom_allocator_test.cpp
+    tdigest_iterator_test.cpp
 )
 
 if (SERDE_COMPAT)
diff --git a/tdigest/test/tdigest_iterator_test.cpp b/tdigest/test/tdigest_iterator_test.cpp
new file mode 100644
index 00000000..e7c03205
--- /dev/null
+++ b/tdigest/test/tdigest_iterator_test.cpp
@@ -0,0 +1,274 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+#include <catch2/catch.hpp>
+#include <memory>
+#include <map>
+#include <vector>
+#include <set>
+
+#include "tdigest.hpp"
+
+namespace datasketches {
+
+TEST_CASE("tdigest iterator: basic iteration", "[tdigest]") {
+  tdigest_double td(100);
+  
+  // Insert 10 distinct values
+  for (int i = 0; i < 10; i++) {
+    td.update(static_cast<double>(i));
+  }
+  
+  // Collect all centroids via iteration
+  std::map<double, uint64_t> centroids;
+  for (const auto&& centroid : td) {
+    centroids[centroid.first] = centroid.second;
+  }
+  
+  // Should have collected all 10 distinct values
+  REQUIRE(centroids.size() == 10);
+  
+  // Verify each value was captured correctly
+  for (int i = 0; i < 10; i++) {
+    REQUIRE(centroids.count(static_cast<double>(i)) == 1);
+    REQUIRE(centroids[static_cast<double>(i)] == 1);
+  }
+}
+
+TEST_CASE("tdigest iterator: explicit begin/end with unique_ptr", "[tdigest]") {
+  // This test reproduces the bug scenario found in ClickHouse
+  std::unique_ptr<tdigest_double> td(new tdigest_double(100));
+  
+  // Insert distinct values
+  for (int i = 0; i < 10; i++) {
+    td->update(static_cast<double>(i));
+  }
+  
+  // Use explicit begin/end iterators
+  auto it = td->begin();
+  auto end_it = td->end();
+  
+  std::vector<double> means;
+  std::vector<uint64_t> weights;
+  
+  while (it != end_it) {
+    // Before the fix, accessing it->first would return garbage or same value repeatedly
+    double mean = it->first;
+    uint64_t weight = it->second;
+    means.push_back(mean);
+    weights.push_back(weight);
+    ++it;
+  }
+  
+  // Should have collected 10 centroids
+  REQUIRE(means.size() == 10);
+  REQUIRE(weights.size() == 10);
+  
+  // All means should be distinct (not all zeros or garbage)
+  std::set<double> unique_means(means.begin(), means.end());
+  REQUIRE(unique_means.size() == 10);
+  
+  // Verify all expected values are present
+  for (int i = 0; i < 10; i++) {
+    REQUIRE(unique_means.count(static_cast<double>(i)) == 1);
+  }
+}
+
+TEST_CASE("tdigest iterator: structured bindings", "[tdigest]") {
+  tdigest_double td(100);
+  
+  for (int i = 0; i < 5; i++) {
+    td.update(static_cast<double>(i * 10));
+  }
+  
+  std::vector<std::pair<double, uint64_t>> collected;
+  
+  // Test structured bindings
+  for (auto it = td.begin(); it != td.end(); ++it) {
+    const auto& centroid = *it;
+    collected.emplace_back(centroid.first, centroid.second);
+  }
+  
+  REQUIRE(collected.size() == 5);
+  
+  // Verify distinct values were collected
+  std::set<double> means;
+  for (const auto& pair : collected) {
+    means.insert(pair.first);
+    REQUIRE(pair.second == 1);  // Each value inserted once
+  }
+  
+  REQUIRE(means.size() == 5);
+  for (int i = 0; i < 5; i++) {
+    REQUIRE(means.count(static_cast<double>(i * 10)) == 1);
+  }
+}
+
+TEST_CASE("tdigest iterator: operator-> access", "[tdigest]") {
+  tdigest_double td(100);
+  
+  // Insert values
+  for (int i = 1; i <= 10; i++) {
+    td.update(static_cast<double>(i * i));  // 1, 4, 9, 16, 25, 36, 49, 64, 81, 100
+  }
+  
+  // Access via operator->
+  std::map<double, uint64_t> centroids;
+  auto end_it = td.end();
+  for (auto it = td.begin(); it != end_it; ++it) {
+    // operator-> should return valid values
+    centroids[it->first] = it->second;
+  }
+  
+  REQUIRE(centroids.size() == 10);
+  
+  // Verify the squared values
+  for (int i = 1; i <= 10; i++) {
+    double expected = static_cast<double>(i * i);
+    REQUIRE(centroids.count(expected) == 1);
+  }
+}
+
+TEST_CASE("tdigest iterator: range-based for with const auto&&", "[tdigest]") {
+  tdigest_double td(100);
+  
+  // Insert values
+  for (double d = 0.0; d < 10.0; d += 1.0) {
+    td.update(d);
+  }
+  
+  size_t count = 0;
+  std::set<double> seen_means;
+  
+  // This pattern was working in simple tests but failing in optimized builds
+  for (const auto&& centroid : td) {
+    seen_means.insert(centroid.first);
+    count++;
+  }
+  
+  REQUIRE(count == 10);
+  REQUIRE(seen_means.size() == 10);
+  
+  // Verify all values from 0 to 9 are present
+  for (int i = 0; i < 10; i++) {
+    REQUIRE(seen_means.count(static_cast<double>(i)) == 1);
+  }
+}
+
+TEST_CASE("tdigest iterator: copy vs reference semantics", "[tdigest]") {
+  tdigest_double td(100);
+  
+  td.update(1.0);
+  td.update(2.0);
+  td.update(3.0);
+  
+  auto it = td.begin();
+  
+  // Store the pair
+  auto pair1 = *it;
+  double mean1 = pair1.first;
+  
+  ++it;
+  
+  // Store another pair
+  auto pair2 = *it;
+  double mean2 = pair2.first;
+  
+  ++it;
+  
+  auto pair3 = *it;
+  double mean3 = pair3.first;
+  
+  // All three means should be distinct
+  REQUIRE(mean1 != mean2);
+  REQUIRE(mean2 != mean3);
+  REQUIRE(mean1 != mean3);
+  
+  // And they should match our input values
+  std::set<double> means = {mean1, mean2, mean3};
+  REQUIRE(means.count(1.0) == 1);
+  REQUIRE(means.count(2.0) == 1);
+  REQUIRE(means.count(3.0) == 1);
+}
+
+TEST_CASE("tdigest iterator: empty sketch", "[tdigest]") {
+  tdigest_double td(100);
+  
+  // Empty sketch should have begin() == end()
+  REQUIRE(td.begin() == td.end());
+  
+  // Range-based for should not execute
+  size_t count = 0;
+  for (const auto&& centroid : td) {
+    (void)centroid;  // Silence unused warning
+    count++;
+  }
+  REQUIRE(count == 0);
+}
+
+TEST_CASE("tdigest iterator: single value", "[tdigest]") {
+  tdigest_double td(100);
+  td.update(42.0);
+  
+  size_t count = 0;
+  double captured_mean = 0.0;
+  uint64_t captured_weight = 0;
+  
+  for (const auto&& centroid : td) {
+    captured_mean = centroid.first;
+    captured_weight = centroid.second;
+    count++;
+  }
+  
+  REQUIRE(count == 1);
+  REQUIRE(captured_mean == 42.0);
+  REQUIRE(captured_weight == 1);
+}
+
+TEST_CASE("tdigest iterator: large dataset", "[tdigest]") {
+  tdigest_double td(100);
+  
+  // Insert 1000 distinct values
+  for (int i = 0; i < 1000; i++) {
+    td.update(static_cast<double>(i));
+  }
+  
+  // Iterator should provide compressed centroids (not all 1000)
+  size_t centroid_count = 0;
+  std::set<double> unique_means;
+  uint64_t total_weight = 0;
+  
+  for (const auto&& centroid : td) {
+    unique_means.insert(centroid.first);
+    total_weight += centroid.second;
+    centroid_count++;
+  }
+  
+  // Should have fewer centroids than input values due to compression
+  REQUIRE(centroid_count < 1000);
+  REQUIRE(centroid_count > 0);
+  
+  // Total weight should equal number of input values
+  REQUIRE(total_weight == 1000);
+  
+  // All means should be unique (no duplicates)
+  REQUIRE(unique_means.size() == centroid_count);
+}
+
+} // namespace datasketches